Measurement: Sensors (Dec 2022)
An enhanced approach for leaf disease identification and classification using deep learning techniques
Abstract
Agriculture isn't just a profession for public yet in adding up a method of life. Most Customs and Cultures in the World spin around Agriculture. Advanced agriculture can possibly make agriculture more gainful, more reliable and to utilize time and assets all the more productively. This brings basic favorable circumstances for farmers and more extensive social advantages far and wide. Identification of plant infection is troublesome for farmers. In the event that identified infection is inaccurate, at that point there is an immense misfortune on the creation of crops and efficient estimation of market. This work introduces new software to help the farmers to identify the infection automatically and offer solutions to the identified problems through expert database. It identifies infected spots of leaf using Fuzzy C-means clustering (FCM) algorithm. Features are extracted using Scale Invariant Feature Transform (SIFT). Long Short Term Memory Networks (LSTM) is used to classify the extracted features. The investigational results are performed on Kaggle Dataset. The projected work is tested for some fungi and bacteria affected leaf of disease classes including Leaf Rust, Blights, Powderly Mildew, Mealy bugs, Downy Angular leaf spot, Mildew, and healthy leaves. The disease classification results illustrate that the projected work outperforms with enhanced accuracy, precision, Recall and F1-Score when compared existing methods. The proposed work tested for around 8000 images, which contains 5278 healthy leaf images and 2725 diseased leaf images achieved up to 96% of accuracy.
